Using soft computing to forecast the strength of concrete utilized with sustainable natural fiber reinforced polymer composites

Abstract

The urgent necessity to strengthen structures with substandard designs has been demonstrated by recent earthquakes. Natural fiber reinforced polymers (NFRPs) provide an affordable, sustainable means of reinforcement, yet accurately forecasting their performance is still a difficult task. The application of soft computing approaches to forecast the compressive strength (CS) of concrete specimens reinforced through various NFRPs is examined in this work. In the present study, three approaches were utilised: AdaBoost, Random Forest (RF), and XGBoost. To evaluate the performance of each soft computing technique, several statistical indicators were calculated, including the Coefficient of Determination (R²), Nash–Sutcliffe Efficiency (NSE), Root Mean Square Error (RMSE), Wilmott Index (WI), Mean Absolute Error (MAE) and Mean Squared Error (MSE). The results demonstrated that the XGBoost model outperformed the other models, with an R² of 0.85, RMSE of 5.05, MAE of 3.83, MSE of 25.48, WI of 0.96, and NSE of 0.85 during the testing stage. SHAP analysis revealed that the unconfined CS of the concrete specimen (fc) had the greatest impact on Forecasting the CS of NFRP. These findings suggest that soft computing has considerable potential to forecast the CS of concrete reinforced utilising NFRPs. XGBoost is a model that generates the most precise forecasts out of all the others, making it an essential tool for engineers who aim to improve the performance and design of structures constructed of sustainable materials.